We built the financial-crime investigation software Poland's law enforcement relies on - now expanding worldwide

Client

Matic is a Polish IT company that has spent decades building technology for public administration, uniformed services, and the defense sector - IT infrastructure, cybersecurity, and law enforcement forensics among its core lines. Financial Trail is its investigative data-analysis product: in use across district and regional prosecutor's offices, provincial and national police headquarters, the Central Cybercrime Bureau, the Internal Security Agency (ABW), the Central Bureau of Investigation (CBŚP), the National Revenue Administration (KAS), and Poland's financial intelligence unit. Matic is now taking it to buyers outside Poland.

Challenge

An analyst opens a money-laundering case and finds this on their desk: 35 bank accounts - personal and corporate - spanning two years, exported as 350 separate text files. No two banks describe the same event the same way; a cash withdrawal reads as a different transaction type depending on which bank issued it.

Currency conversions arrive with no exchange rate attached. Some entries are outright wrong - a transfer recipient logged under a joke name because a teller typed it that way. Multiply that across roughly 1.5 million individual transaction records, and the analyst's actual job - who moved money to whom, and does it add up to a crime - can't start until all of that is untangled.

Financial Trail existed before Pirxey got involved, but not in a shape anyone could build on. An earlier version had shipped years before; a second rebuild followed it and stalled - hard to extend, and not something you could point at a million-row dataset with any confidence. Matic started fresh on a third generation and brought Pirxey in as engineering partner from the start.

The brief wasn't "build an app that displays a table." It was: take data that is simultaneously enormous in volume and specific in meaning case by case, and make the platform handle both scales at once - fast enough for an analyst to work interactively, not submit a batch job and check back the next morning.

What we built

The foundation is a normalization layer: every import, regardless of source bank or file format, resolves into one consistent schema, where a sender is always a sender and a withdrawal is always a withdrawal.

Duplicate or misspelled entities collapse into a single tagged identity instead of staying scattered across a dozen near-matches. Investigators configure that matching logic themselves - exact, approximate, "contains," or "excludes" - against reference databases they build or import from CSV, and can set the system to re-tag automatically as new data lands, or re-verify existing tags whenever a record changes.

On top of that sits the analysis layer: pivot views, balance calculations, path-tracing across long money trails, and a visual query builder that lets an analyst - not a developer - construct relationship queries across transactions, invoices, identities, and bank accounts without writing anything resembling code, then save and re-run those patterns on the next case.

None of it works if it's slow. The constraint that shaped almost every decision was speed at scale: a filter, sort, or graph render had to return in seconds against a database of well over a million records, not the batch-job timelines the previous version was stuck with. That bar is now written directly into Matic's own specification for the current build phase: tag matches across a database of 1,000,000+ records in under 30 seconds, and export or import saved queries and reference databases - encrypted, password-protected - between cases and investigating teams.

None of this is domain-neutral engineering. Building a shopping cart is intuitive to any developer; building the semantics of who legally owns a joint bank account in Poland, or why the same ATM withdrawal shows up differently depending on which bank processed it, is not. Getting those decisions right meant learning the domain well enough to apply it at every layer of the product, not just the parts a user sees.

Outcome

The original engagement had a defined scope and end date. It ended on schedule, and instead of closing out, Matic signed Pirxey to a standing collaboration - the same team is still building the product, well past the original timeline.

Financial Trail now runs inside institutions spanning the full chain of Poland's financial-crime enforcement, from district prosecutors to the national financial intelligence unit, and Matic is actively selling it abroad.